Core offer
SuperCompress is open-source prompt compression for AI apps. It cuts oversized LLM inputs before OpenAI, Claude, Gemini, or local models see them.
Creator kit
Campaign assets for selling prompt compression without sounding like an ad.
Use these angles, scripts, and templates with your own referral link. The best content is a quick before/after demo: paste oversized context, compress it, show the tokens saved, then point builders to the free API key.
Affiliate payout
40% recurring. Starter pays $4/mo, Pro pays $8/mo, Business pays $24/mo for each active referred customer.
Creator CTA
Send people to your referral link first, then tell them to create a free API key. No card required for the free tier.
Best campaign angles
| Angle | Hook | Demo |
|---|---|---|
| RAG cost leak | Your RAG app is probably sending useless chunks to the model. | Paste retrieval output, ask a specific query, show before/after tokens. |
| Agent trace cleanup | Agents generate huge tool traces. Most of it should not hit the next LLM call. | Paste a long coding-agent log, compress against the next task. |
| Truncation failure | Head/tail truncation can drop the one line that answers the question. | Compare blind truncation with query-aware compression. |
| API quick win | Add one preprocessing call before your LLM call and measure what you saved. | Use the cURL endpoint with a free key and show the response metadata. |
X / LinkedIn post
Most AI apps are paying for context the model does not need. I tested SuperCompress on a long RAG/chat-history prompt: - original context: [X] tokens - after compression: [Y] tokens - saved: [Z]% It runs before OpenAI, Claude, Gemini, or local models. Open source + free hosted API key. Try it here: [YOUR_REFERRAL_LINK]
Short video script
"Here is a hidden cost in AI apps. This prompt has RAG chunks, chat history, and tool output. It is huge. Before sending it to the model, I run it through SuperCompress. Now the model gets the context that matters for this query, but not all the noise. That means fewer input tokens, less prefill work, and lower LLM cost. It is open source, and the hosted API has a free key. Link is in the description."
Email/newsletter blurb
Tool worth trying: SuperCompress. If you are building with RAG, agents, support logs, or long chat history, you may be sending thousands of unnecessary input tokens to your LLM provider. SuperCompress runs before the model call and compresses context around the current query. It works with OpenAI, Claude, Gemini, LangChain, Vercel AI SDK, and local flows. Free API key: [YOUR_REFERRAL_LINK]
Disclosure language
Use a clear disclosure anywhere you post an affiliate link. Keep it simple:
Disclosure: this is an affiliate link. I may earn a commission if you become a paid SuperCompress customer.
Need your link?
Generate it on the affiliate page, then replace [YOUR_REFERRAL_LINK] in the templates above.